OpenAI strengthens ChatGPT safety for teenagers
OpenAI is tightening ChatGPT’s responses about weapons, mass attacks, and signs of violence, with special measures for users under 18. It is also announcing a community safety report to be published every six months and reporting more than 200 notifications to authorities during the third quarter of 2026.
OpenAI is announcing new safety measures for ChatGPT, with tighter limits on weapons and violence, specific controls for minors, and a regular report on how it detects and responds to potential threats. The company also says it made more than 200 reports to authorities during the third quarter of 2026 about imminent threats of violence.
A report on community safety
OpenAI will publish its first Community Safety Report at the beginning of 2027, focusing on preventing violence and other risks to communities. The document will cover the second half of 2026 and will be updated every six months.
The report will explain how its systems detect concerning behavior, how cases are reviewed, and what measures are applied. It will also include anonymized examples, data on the effectiveness of its protections, and changes to its policies.
The company stresses that AI cannot resolve violent situations on its own. That is why it combines its systems with intervention from authorities, families, mental health professionals, and prevention experts.
Less detail about mass attacks
ChatGPT already prohibits providing instructions for planning or committing violent acts. It will now also limit responses about mass attacks that occurred in the past, even when the person asking does not show dangerous intent.
The model will provide less detail about:
- The weapons and equipment used.
- How those resources were obtained.
- Detailed reconstructions of the attacks.
- Comparisons about the scale of the harm that could increase the attacker’s notoriety.
This does not mean ChatGPT will stop explaining these events. It can still help people understand their impact on victims and communities, identify warning signs, and study how to prevent future incidents.
More restrictions when there are signs of risk
An apparently normal conversation can take on a different meaning when combined with earlier messages. OpenAI is working to make ChatGPT use brief, objective summaries of recent risk signals, so it responds more cautiously when a request could facilitate violence.
The company also wants to apply stricter limits when it estimates that the user is under 18, even when there are no clear signs of violent intent. If a conversation shows a persistent fixation on violence or seriously harming another person, ChatGPT may automatically close the chat to interrupt that pattern.
If this happens several times, the account could be paused temporarily. Accounts that violate the rules on violent activity may also be blocked.
What changes for teenagers
ChatGPT for teenagers already applies additional protections. From now on, minors will receive less practical information about weapons and security systems that could be used to facilitate violence, even when the question appears harmless.
They will still be able to use the service to prevent or de-escalate conflicts, ask for support, and seek help from a trusted adult. The goal is to limit operational information, not prevent teenagers from learning to recognize a dangerous situation or ask for help.
OpenAI has also strengthened its age prediction system, which analyzes signals such as usage patterns to estimate whether a person may be a minor. If it mistakenly places an adult in the teen experience, the adult can confirm their age through Persona, the company’s verification partner, using a process designed to protect their privacy.
Parental controls allow an adult’s account to be linked to their teenager’s account. Families can already receive alerts about certain serious concerns involving self-harm and eating disorder-related behavior. Since July, they can also be notified when the minor’s account is deactivated for violating the rules on violence.
OpenAI is considering adding alerts when a teen account is paused after several conversations related to violence. The new protections will begin rolling out over the coming weeks and will continue to be adjusted over time.
For you, the most visible change will be that ChatGPT may respond with less detail to questions about weapons, previous attacks, or violence when it detects a potential risk, especially if it believes the person asking is a minor. The key question will be how these detections work in practice and whether OpenAI can balance prevention with useful answers for people seeking to understand or avoid violence.